Automated 3D CT Segmentation for Hepatic Steatosis Assessment Using MRS-PDFF

Overview

This study evaluates the clinical utility of fully automated 3D organ segmentation for assessing hepatic steatosis on pre-contrast and post-contrast CT images, using MRS-PDFF as the reference standard. It compares volumetric CT attenuation measurements with traditional ROI-based methods to determine accuracy and potential interchangeability.

Background

Hepatic steatosis, characterized by fat accumulation in hepatocytes, is a hallmark of non-alcoholic fatty liver disease. While liver biopsy remains the gold standard for diagnosis, its invasiveness limits routine use. Non-invasive imaging techniques such as MRI-PDFF and MRS provide accurate quantification of liver fat, but CT, despite its limitations and ionizing radiation exposure, is widely used and cost-effective for opportunistic screening. Recent advances in deep learning have enabled automated 3D segmentation, potentially improving CT-based hepatic steatosis assessment by providing volumetric attenuation measurements.

Data Highlights

CT attenuation values of liver and spleen were measured using two methods: manual ROI-based and fully automated 3D volumetric segmentation on pre-contrast and post-contrast CT images. The study included adults undergoing preoperative liver CT and MRI with MRS-PDFF as the reference. Automated segmentation provided mean liver and spleen attenuation values in Hounsfield units (HU), with an example showing liver HU of 36 (pre-contrast) and 105 (post-contrast), and spleen HU of 55 (pre-contrast) and 147 (post-contrast) in a subject with 25.7% MRS-PDFF.

Key Findings

• Fully automated 3D segmentation enables volumetric CT attenuation measurement of liver and spleen on both pre-contrast and post-contrast images.
• Volumetric CT attenuation measurements may better represent heterogeneous hepatic steatosis compared to ROI-based methods.
• CT attenuation values inversely correlate with hepatic fat content, decreasing as steatosis severity increases.
• Automated volumetric measurements showed potential interchangeability with manual ROI-based measurements for hepatic steatosis assessment.
• Pre-contrast CT attenuation measurements remain preferred, but post-contrast images can also be utilized effectively.

Clinical Implications

Automated 3D segmentation techniques can facilitate objective, comprehensive, and efficient assessment of hepatic steatosis using CT, potentially improving screening and monitoring in clinical practice. This approach may reduce operator dependency and variability inherent in manual ROI-based measurements, supporting broader use of CT as a cost-effective opportunistic tool for liver fat quantification.

Conclusion

Fully automated 3D organ segmentation on CT images provides a reliable and practical method for assessing hepatic steatosis, showing promise as a complementary tool alongside established non-invasive reference standards like MRS-PDFF. This technique may enhance the clinical utility of CT in liver fat evaluation.

References

1. 1 -- Liver biopsy as gold standard for hepatic steatosis
2. 2,3 -- MRI-PDFF and MRS as non-invasive reference standards
3. 4 -- Validation of MRI-PDFF correlation with histology and triglycerides
4. 5 -- Quantitative ultrasound for hepatic steatosis
5. 6 -- CT as cost-effective opportunistic screening tool
6. 7-11 -- CT attenuation metrics for hepatic steatosis assessment
7. 12-14 -- Deep learning and automated segmentation in medical imaging
8. 15 -- Prior methodology for ROI-based CT attenuation measurement